The version I compared against was the one using numeric comparison ( <= ) not ternary operators. It's in the link I shared.
So I hope you revert your downvote or provide proof of what I am saying is incorrect.
You are completely right, I'll mention this in the post (In that case do a branchless comparison based on subtraction or similar?). Fwiw when doing simd, you'd use intrinsics which wouldn't have this issue, but it is a valid point and completely true. Thanks!
"You are right!" + no upvote = bad
I'm not the one who downvoted you, but I am the one who wrote the article :P